E3 Pocket
Private AI workstation for local inference and edge applications
Compact NVIDIA-powered infrastructure for running compatible AI models locally, available in E3 Pocket T4000 and T5000 configurations.
E3 Pocket T4000
Why organisations choose it
Run compatible AI models and assistants on dedicated local hardware, keeping AI workloads close to your organisation's data.
Up to 128GB unified memory and up to 2,070 Sparse FP4 TFLOPs support advanced inference and multimodal AI workloads.
A 130 × 125 × 75 mm desktop footprint brings private AI infrastructure to enterprise, government, research and robotics environments.
Available models
2 models in this family.
W638-A
- Module
- NVIDIA Thor T4000
- Unified memory
- 64GB
- Storage
- 1TB NVMe
W638-B
- Module
- NVIDIA Thor T5000
- Unified memory
- 128GB
- Storage
- 2TB NVMe
| Part number | Module | Unified memory | Storage |
|---|---|---|---|
| W638-A | NVIDIA Thor T4000 | 64GB | 1TB NVMe |
| W638-B | NVIDIA Thor T5000 | 128GB | 2TB NVMe |
Technical specification
Chassis
- Dimensions
- 130 × 125 × 75 mm
- Weight
- Approx. 1.3kg
Connectivity
- Wireless
- Wi-Fi 6, Bluetooth 5.2
- Network
- 2 × 10GbE
- Display
- 2 × HDMI 2.1
Software
- Operating system
- Ubuntu 24.04
Specification notes
Module core counts are confirmed with the selected E3 Pocket configuration.
Discuss this configurationTell us what you want to build.
Share your device requirements, target quantities and market. Our team will help you explore suitable products and the next steps for your project.
- Distributors and resellers taking a range to their own market.
- Education, government and institutional buyers.
- Enterprise buyers and organisations with product requirements.